SEO After AI: What Still Matters, What Changed, and Why

A search result can now answer a question without sending the person to the website that supplied the information. Google may show a page as a source while that page receives no visit. For a business that used to judge SEO by clicks, that’s an uncomfortable change.

Much of the advice businesses have been encouraged to publish is easy to summarize. A checked result, a useful comparison or something learned from doing the work gives somebody a better reason to open the page. Research into changing click behavior matters, but so does the more immediate question: what would a reader gain by visiting your website?

What AI Changed in Search

Google’s generative search features expand the amount of work that can happen between a query and a click. AI Overviews can summarize a topic and provide supporting links. AI Mode is designed for more exploratory questions, comparisons, and multi step research. Google says both features may use query fan out, where the system issues several related searches across subtopics and data sources before assembling a response.

Ask AI Mode which CRM suits a small sales team and its answer may draw on sources about price or migration, even though neither was mentioned in the question. That branching process is part of query fan out. A page covering one of those narrower concerns can become relevant without repeating the original wording.

Google’s AI features documentation describes those experiences as part of Search, drawing on familiar systems for finding and ranking useful information. It doesn’t offer a separate shortcut that guarantees an appearance in an AI answer.

Search Visibility and Website Traffic Are Separating

More visibility does not always mean more visits. If a guide begins appearing more often in Google’s AI results while its organic clicks stay flat, check which URLs are appearing and how the broader search query mix is changing before calling it a win or a loss. An impression shows exposure; it doesn’t tell you whether that person later contacted the business.

The effect on clicks varies by query and result type. Pew Research Center’s analysis of U.S. browsing behavior in March 2025 found that users clicked a traditional search result in 8% of visits where an AI summary appeared, compared with 15% of visits without one. Ahrefs reported a separate observational study in February 2026 in which the top ranking page had a substantially lower average click through rate when an AI Overview appeared. The studies use different datasets and methodologies, so the percentages are best read as evidence of a directional shift in click behavior. Pew’s study and Ahrefs’ 2026 update provide the underlying context.

Commercial search is part of this shift as well. Semrush analyzed more than 600,000 U.S. desktop keywords from November 2025 through April 2026 and found AI Overviews increasingly appearing alongside advertising and commercial intent searches. The practical implication is that generative search belongs in the full customer journey, including research and comparison stages that can precede a purchase or inquiry. Semrush’s study explains the methodology and limits.

How AI Search Fits Into the SEO Stack

Google describes its generative search features as grounded in the Search index and core ranking systems. Retrieval augmented generation can use those systems to find relevant current pages, while query fan out expands one user request into related searches. The generated response then presents supporting links alongside synthesized information.

For Google’s AI Overviews and AI Mode to show a page as a supporting link, Google says the page must be indexed and eligible to appear in Search with a snippet. That puts ordinary indexing problems ahead of specialized AI tactics. If Google is choosing another URL as canonical, investigate that before rewriting the article for an AI answer.

Our Retrieval Layer guide goes deeper into how search systems choose which evidence reaches a model. Here, the concern is what website owners can do when their useful pages aren’t being found or cited.

Do You Need GEO or AEO?

AEO and GEO are useful labels when they help teams describe a measurement problem or a search surface. Google now states explicitly that, from its perspective, optimization for generative AI Search is still SEO because AI Overviews and AI Mode depend on the same underlying Search systems.

If research articles gain generative AI impressions without bringing more qualified inquiries, check which pages appeared and whether their topics answer questions future customers ask. Then look at what a visitor can do on those pages. Publishing more articles labeled “AI optimized” won’t explain why the existing ones aren’t producing useful enquiries.

Our GEO guide explains how to check whether a brand appears in AI answers and whether those appearances lead anywhere useful. That reporting can sit alongside the rest of your SEO work.

Why SEO Still Matters After AI

Traditional SEO still matters after AI Mode because Google’s AI search features depend on pages it can find and use. A page that isn’t indexed can’t appear as a supporting link in an AI Overview or AI Mode, and a page that repeats common advice offers little reason to click through. Fixing accessibility, giving each URL a clear purpose and publishing verifiable material remain sensible priorities for ordinary search too.

AI also makes weak publishing strategies easier to scale. A team can generate hundreds of articles quickly, which increases the importance of editorial judgment. Google warns against producing large numbers of pages for query variations primarily to manipulate rankings or generative responses. Distinct page ownership and genuinely useful coverage matter more when content production itself is cheap.

Where AI Automation Helps an SEO Team

AI can save time on the work of sorting and comparing information. It can group similar Search Console queries, flag repeated titles, compare page versions and suggest related articles. Those suggestions still have to be checked against the source data and the way the website is organized.

A spreadsheet cannot tell the whole story about which services bring the right customers or which technical change could disrupt a working page. Someone who knows the business needs to set priorities and approve changes before they go live. The useful applications are fairly specific.

For a recommendation that could change search results, record the affected URLs, the evidence behind the change, who approved it and what will be checked afterward. A quick suggestion is useful only when the team can test it safely and understand the result.

What Matters More Now

Distinct page ownership

Each important URL should have a clear search job. A broad guide can own the main topic while supporting pages handle narrower problems, comparisons, or applications. This reduces internal overlap and gives both users and search systems a cleaner path through the subject. Our guide to website structure for SEO at scale explains how to build that hierarchy.

Distinctive information

Google’s 2026 guidance encourages original material grounded in experience. A technical article could show how a booking integration behaves when the other service goes offline, including how the writer checked whether the request arrived. Even a short, well documented test can teach more than a page full of familiar recommendations.

Before publishing, read the strongest existing explanations of the subject. Where does yours help somebody further? If nothing stands out, talk to the person who does the work before adding more paragraphs. Ask what they check first when the system fails and what would convince them it was repaired.

Internal architecture

Internal links continue to help Google discover pages and understand relationships. They also become increasingly important when a site publishes across adjacent AI, SEO, technical, and commercial topics. The links should express a real relationship between pages and reinforce which page acts as the main reference for a topic.

Authority and external evidence

Relevant links, citations, reviews, third party coverage, expert references, and other external signals help establish that a site exists within a real information ecosystem. Google’s guidance also warns against manufacturing mentions for generative search. Durable authority comes from material that other people have a reason to cite, discuss, recommend, or verify.

External authority still deserves its own analysis within the wider search strategy. Our guide to how backlinks affect SEO rankings covers that part in more detail.

Multimodal usefulness

Google’s current AI search guidance encourages useful images and video alongside text where they improve the page. A diagram, annotated screenshot, demonstration, comparison table, or short video can answer parts of a question that prose handles poorly. Images and video should add evidence or make the page easier to use, with ordinary image and video SEO practices applied correctly.

SEO Tactics That Aren’t Worth Chasing

Google’s current guidance resolves several questions that attracted heavy speculation during the first years of generative search. Teams can use that guidance to avoid spending time on unsupported tactics.

• AI specific text files: Google says llms.txt has no positive or negative effect on visibility or rankings in Google Search. Other services may choose to use it, so the decision depends on the systems a site actually targets.
• Artificial content chunking: Google says its systems can understand nuanced topics across a normal page, so section length should follow the reader’s needs and the subject.
• AI only rewrites: Google can connect queries with synonyms and related meanings, which reduces the value of rewriting a page solely to cover every phrasing variation.
• Special generative search schema: Google uses the existing Search eligibility framework for AI Overviews and AI Mode. Structured data remains useful when it accurately represents visible page content and supports ordinary rich result eligibility.
• Mass pages for fan out variants: Google’s guidance warns against creating many pages around possible query variations primarily to manipulate Search or generative responses.
• Manufactured mentions: Google explicitly advises against chasing inauthentic mentions as an AI visibility tactic.

A business with an important service page Google hasn’t indexed and twenty ideas for new AI search articles should investigate the existing page first. Its canonical, internal links and content are concrete things to check. Publishing another batch of speculative pages leaves the original problem untouched and gives the team more content to maintain.

Give People a Reason to Visit Anyway

An AI answer can explain that a customer relationship management system connects with a booking form. It can’t tell a business whether that connection reliably saves every inquiry, sends it to the right person or loses the information needed to call the customer back. Someone has to test those details. A guide that shows what to check, what can fail and how to confirm it worked offers something an ordinary summary often won’t.

That’s where I’d spend more of a content budget. Publish findings from tests you can explain. Answer the awkward question a sales page would rather avoid. Show what a service includes, what it excludes and what changes the price. If two options are both good, explain who should choose each one. Even a small, carefully checked example can be more useful than a long roundup of general advice.

It also takes restraint. Don’t invent numbers, testimonials or experiences to make a page feel original. A careful explanation of what you have and haven’t verified earns more trust than a dramatic story that never happened.

The systems behind AI answers can still miss good evidence or pull the wrong part of a source. Our guide to retrieval and search systems explains some of those limits for readers who want the technical detail.

How to Measure SEO After AI

Google now gives site owners a direct view into generative search visibility. Its Search Console Generative AI performance reports rolled out globally by August 31, 2026 and show how often site URLs appear in generative AI features, which pages appear, the countries involved, device data for Search, and performance over time. Google’s announcement and documentation provide the reporting definitions.

Those impressions should sit alongside ordinary Search Console and analytics data. A practical measurement framework separates visibility, traffic, and business impact so each metric is interpreted within the stage it actually measures.

AI visibility data needs context. An impression shows that one of your URLs appeared in a generative feature; it doesn’t prove that the person visited your website or later became a customer. Google’s report documentation also notes that a site may not see the separate report if it hasn’t received enough qualifying impressions. Use the report alongside ordinary search data, landing-page behavior and customer feedback rather than treating impressions as sales.

What I Would Change on a Real Website

Before buying another batch of articles, open the pages that already attract visitors. Read each one with the search results beside it. If the page tells readers little more than the AI summary, why would they click through? It may need an explanation drawn from the work or a straightforward answer about the service’s cost.

Start with a few pages that already match real search demand. Give readers information they couldn’t find in the results preview, and link to another guide only when it helps them finish what they came for. Compare the business enquiries with the search reports instead of treating an appearance in an AI answer as a conversion.

Google’s own advice for generative AI search now favors original, useful material over special markup tricks or pages made for every possible search variation. That’s a sound editorial test too. If the article adds nothing that a reader couldn’t get from ten similar pages, it needs more thought before it needs more promotion.

The Next Shift: Search Agents That Can Act

Generative answers are already changing discovery, and agentic systems add another layer: software that can inspect a site and complete a task for the user. Google’s 2026 guidance points website owners toward emerging agent friendly practices and protocols such as Universal Commerce Protocol for experiences where Search agents may compare products or complete actions.

For a store or booking service, being found is only the beginning. A search agent may also need to identify the right product or check whether an appointment is available. The website needs accurate information and a reliable way to complete that interaction. A broken booking flow is still a broken booking flow, however sophisticated the agent becomes.

Before rebuilding a site for emerging search agents, check whether its existing pages accurately describe what people can buy or book, and whether the forms still work. That is worth fixing regardless of which new search features last.